Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Advanced Micro Devices - Moderated (AMD) -- Ignore unavailable to you. Want to Upgrade?

To: KeithDust2000 who wrote (156616)	4/17/2005 1:26:00 AM
From: pgerassi	Read Replies (2) \| Respond to of 275872

Dear Keith: I looked at Xeon DP 2MB L2 at 3.6GHz and it loses to a Opteron 252. It would still lose, if it had increased clock to 3.8GHz. It still loses, if it also got double the L2. It still loses, if it had a 1066MHz FSB. FBDIMMs would cause it to lose even more due to the increased latency. Now by the time it and the chipset it needs to do well comes out, where will AMD be at? Opteron 280 will be out by Q4, possibly even Q3 of this year. More than likely it will be at least to Opteron 285 (2.6GHz by 2) on 90nm. It may be on a new socket on 65nm by then as Fab 36 is operational. A 2.6GHz Opteron DC with 2MB L2 each and 4 channel DDR2 (maybe DDR3) on die with 4 HT3 links, one of which is cHT (for 4xx having 2 cHTs, 8xx having 3 cHT and 16xx having all 4 being cHT links). We know Socket M2 is coming for desktops and mobiles, why not one for servers called M1 having the above. HT1 went to 800MHz, HT2 goes to 1.4GHz so HT3 must go to at least 2GHz. Dempsey could be old and way too slow by the time it comes out. Like the (I forget what it was called) P4 that was supposed to replace P4 this year, it could get cancelled. It seems a lot of Intel designs have that fate. A lot of them also miss their targets and underwhelm when they see the light of day. Don't fall into the trap of comparing a future CPU with on paper capabilities to a currently existing CPU. All sorts of things may happen, it could slip (many do and it would not be that unusual), it could have its targets reduced (just look at Prescott for an example), it could get cancelled (the 4GHz P4 comes immediately to mind) and it could turn out to be a flop (Merced, RDRAM and MTH come to mind). Look at it by comparing IA32 only 3GHz NW Xeon MP at 400MHz FSB and 2MB L3 versus Opteron 875. It is a slaughter. Intel has to hit a moving target and it is not very nimble even when it was ahead. Its trying to push an untried new socket, an untried new process and an untried new CPU where it has recently been burned before as a accomplished fact. Lets look at the PR used to promote the Prescott. It was to be at 4GHz in 2003, have higher IPC than NW, have an awesome 1MB L2 and run much cooler to boot. It only accomplished 1 out of those 4. One target will be never achieved, 4GHz. It missed by three quarters even coming out and will never be considered to have better IPC or lower power than NW. Pete

To: KeithDust2000 who wrote (156616)	4/17/2005 2:39:54 AM
From: dougSF30	Respond to of 275872

Oh come on Keith. AMD can deliver the same performance as Dempsey for 60% of the power. If AMD wanted a 150W TDP, they could blow Dempsey away.

To: KeithDust2000 who wrote (156616)	4/17/2005 8:57:14 AM
From: TechieGuy-alt	Read Replies (1) \| Respond to of 275872

TG, have you looked at Dempsey (65nm Xeon DP)? 3.8Ghz, 4MB L2, Dual FSB @ 1066Mhz (per core bandwidth increases over current Xeon), FBDIMM. Knowing how well Xeon DP 2M @ 3.6Ghz competes against Opteron @ 2.6Ghz, it´s easy to see how DC Opteron 280 (2.4Ghz) will likely compare to 3.8 Ghz Dempsey. It´s a brute force approach, but it should be effective, and one reason why we´re not seeing more interest in Opteron from OEMs. What can AMD do? I'm not sure. Are these dempsey samples in OEM hands already? Otherwise no way they can make Q1 06. How is the 65nm process? Power? Speed? Yields? How much will that platform cost? One thing that they could do is enable a quad DC platform (8 true cores) for the same price as a dual Dempsey (4 cores). No way Dempsey can compete with 8 Opteron cores in one box. There is no chipset penalty to doing that and in terms of price AMD can easily charge 1/2 of what Intel will probably charge for Dempsey to make lots of money. Plus, that'll suddenly make that "regular" (2P, 4P) Xeon market completely uncompetitive. TG